The immune landscape in brain metastasis

Abstract

The prognosis for patients with brain metastasis remains dismal despite intensive therapy including surgical resection, radiotherapy, chemo-, targeted, and immunotherapy. Thus, there is a high medical need for new therapeutic options. Recent advances employing high-throughput and spatially resolved single-cell analyses have provided unprecedented insights into the composition and phenotypes of the diverse immune cells in the metastatic brain, revealing a unique immune landscape starkly different from that of primary brain tumors or other metastatic sites. This review summarizes the current evidence on the composition and phenotypes of the most prominent immune cells in the brain metastatic niche, along with their dynamic interactions with metastatic tumor cells and each other. As the most abundant immune cell types in this niche, we explore in detail the phenotypic heterogeneity and functional plasticity of tumor-associated macrophages, including both resident microglia and monocyte-derived macrophages, as well as the T-cell compartment. We also review preclinical and clinical trials evaluating the therapeutic potential of targeting the immune microenvironment in brain metastasis. Given the substantial evidence highlighting a significant role of the immune microenvironmental niche in brain metastasis pathogenesis, a comprehensive understanding of the key molecular and cellular factors within this niche holds great promise for developing novel therapeutic approaches as well as innovative combinatory treatment strategies for brain metastasis.

Brain metastases (BrM) arise in around 25% of all adult cancer patients, constituting the most common brain neoplasm in adults with the highest prevalence in nonsmall cell lung cancer (NSCLC), breast cancer, and melanoma.¹ Despite intensive therapy including surgical resection,² radiotherapy,³ and drug treatment,⁴ patient survival rarely exceeds 2 years.⁵ Beyond cancer cell-intrinsic mechanisms, the tumor microenvironment (TME) comprising stromal and immune cells has become the focus of intensive investigation in systemic, as well as central nervous system (CNS) tumors.⁶ The significance of the brain metastatic microenvironment (brain-MME) has been demonstrated in several studies, showing that successful brain colonization by disseminated cancer cells depends on tumor cell–immune–stroma interactions.⁶ The heterogeneous, immunosuppressive nature of the brain-MME has been linked to therapeutic failure,⁷ and immune microenvironment signatures correlate with the prognosis of patients with BrM.⁸

The brain consists of a unique cellular landscape exhibiting remarkable differences in comparison to other tissues frequently associated with metastasis. Beyond a diversity of interconnected neurons, the brain architecture is characterized by the presence of glia cells such as astrocytes and oligodendrocytes, pericytes, ependymal cells, and endothelial cells, as well as a unique extracellular matrix.⁶ The role of these stromal components of the brain-MME, including astrocytes, neurons, and the blood–brain barrier (BBB), has been recently reviewed elsewhere.⁹

The brain-MME is defined by a rich diversity of immune cells. Historically assumed to be immune privileged due to the isolation from peripheral immune cells by the BBB, research over the past years has challenged this narrative by revealing neuroimmune interactions in health and disease. This is particularly true in the context of brain tumors that are accompanied by a compromised BBB allowing for a substantial influx of peripheral immune cells. Consequently, the immune brain-MME is not only characterized by resident microglia, macrophages, and dendritic cells (DCs) but also by infiltrating monocytes, blood-derived macrophages, neutrophils, and lymphocytes (Figure 1).

Myeloid Cells

As an integral part of both innate and adaptive systemic immunity, myeloid cells including macrophages, neutrophils and DCs are increasingly recognized as a key factor in brain physiology and as central regulators of metastasis.

Tumor-Associated Macrophages

Tumor-associated macrophages (TAMs) constitute one of the most abundant cell types found in the brain-MME, accounting for around 30%–40% of nontumor cells.^10–12 TAMs comprise 2 subpopulations defined by tissue origin: CNS-resident TAMs and infiltrating, monocyte-derived macrophages (MDMs).⁶

CNS-resident TAMs can be subdivided into 2 populations: the predominant tissue-resident macrophages microglia, and a group of macrophages populating the CNS borders (choroid plexus, perivascular spaces, meninges), collectively termed border-associated macrophages (BAMs). Microglia originate from yolk-sac–derived progenitors during embryological development and are maintained by self-renewal rather than peripheral monocyte influx. Functioning as the brain’s primary immune cells, microglia are paramount for local immune defence owing to a diverse set of receptors including chemokine, pattern recognition, and purinergic receptors, termed the microglia sensome. Microglia possess substantial phenotypic plasticity, as their transcriptional differences have been described in relation to regional differences in the brain.¹³ Also stemming from yolk sac as well as fetal liver origins,¹⁴ BAMs serve as the tissue-resident macrophagic population at the CNS border regions. BAMs have been described to contain increased antigen presentation capacity compared with microglia, potentially due to localizing at brain–periphery interfaces.¹⁴ Notwithstanding, little is known about the selective contribution of BAMs in BrM,⁷ and thus far, microglia are considered the primary source of CNS-resident TAMs.

Monocytes, not physiologically detected in the brain besides small populations residing in the meninges, are recruited to the brain through chemokine release during BrM⁹ and differentiate into MDMs intratumorally.⁶

Heterogeneity of TAMs

.—Understanding the functional contributions of TAM subpopulations by ontogeny has been challenged with difficulty in distinguishing TAM subsets.^9,15 Classical markers used to differentiate microglia from MDMs such as CX3CR1, CD45, P2RY12, and TMEM119 fail to reliably delineate the two populations, particularly after immune-shaping activities of metastatic cells.^16,17

Extending the identification of subsets beyond the use of lineage tracing and conditional genomic modifications in mouse models, several markers have been proposed for better cell separation over the last years. For example, MDMs have been suggested to be identifiable by the expression of CD49D (ITGA4),¹⁶ CD141, and ICAM.¹⁰ Discrepancy concerning the regional distribution of TAM subsets in BrM likely reflects the longstanding difficulty to finely disentangle resident microglia from infiltrating MDMs. Employing ITGA4 as an MDM marker revealed that microglia-derived TAMs predominantly populate the tumor margin while being largely absent from the tumor core. MDMs in turn were detected primarily in the tumor core, and special subsets of MDMs expressing markers of antigen presentation are localized in the proximity of tumor blood vessels.¹⁰

Additionally, TAM composition is highly dynamic during the course of disease. In early, small BrM lesions, the TAM compartment mainly consists of microglia, continued influx of monocytes over disease progression shifting the composition towards a predominance of MDMs.^18,19 Notably, the described monocytic influx occurs independently of therapeutic disruptions of the BBB and rather constitutes a BrM inherent feature.¹⁶

Phenotypic plasticity of TAMs in BrM

.—In line with the phenotypic variance of microglia and macrophages in health and disease, TAMs in BrM exhibit several notable phenotypic characteristics. In general, education by the brain-MME induces stark phenotypic convergence of TAMs independent of ontogeny.¹⁷ Although there is no clear pattern of a traditionally defined M1 or M2 phenotype,¹¹ the pro- and anti-inflammatory signatures of TAMs in BrM have been comprehensively described and correlated with clinical outcome. For example, an inflammation-associated IFNγ signature is correlated with prolonged survival in lung cancer BrM patients.²⁰

BrM-associated microglia show an activated stellate and reactive amebic morphology throughout the course of disease.^10,18,21 Beyond an observed downregulation of the homeostatic microglial markers CX3CR1, P2RY12, and TMEM119, microglia in the brain-MME show little similarity to other neurological disease-associated microglia.¹⁷ Although microglial phenotype generally does not align well with a reductionist M1-M2 or anti-/pro-inflammatory dichotomy,⁹ BrM-associated microglia population display an immature and inflammatory phenotype,²² with an induced expression of pro-inflammatory cytokines, interleukins, and TNF, as well as the complement cascade and S100 family proteins.^11,17,18,23 Besides, enrichment of exosomal genes, TYRO3/AXL/MER receptor signaling, and lipid metabolism have also been reported.^18,23 Increased gene expression of innate immune sensing genes such as Toll-like receptor signaling and activation markers (CD14, CD163, and CD64) has been described.¹⁰ Notably, upregulated expression of several anti-inflammatory genes has also been associated with microglia in BrM, including LGALS1/3 and IFITM3,¹⁷ VEGFA/B, IGF1, HGF, and FGF20.²⁴

BrM-associated MDMs constitute a more transcriptionally heterogeneous group inclusive of subsets such as microglia-similar, phagocytic/lipid-, hypoxic-glycolytic macrophages, or those expressing anti-inflammatory or IFN signatures.⁷ This diversity of MDM subsets could be reflected in functional differences regarding immunological capacity and tumor support.⁷ Interactions of MDMs with tumor cells and the brain-MME can induce downregulation of CCR2, FCN1, and VCAN while upregulating macrophage markers such as CD163, MERTK, and C1QA during the differentiation of monocytes into macrophages.^10,17 In contrast to microglia, BrM-MDMs display more induced markers of the alternative activation (M2-like) phenotype including CCL17, CCL22, CD163, and TLR1.¹⁸ Several anti-inflammatory molecules, such as FCGR2B, CLEC10A, ARG1, CD39, TGFβ, IL10, IL1RN, CD155, MSR1, and MHC-II, are abundantly expressed in MDMs.^9,11,16 Consequently, BrM-MDMs are frequently associated with antigen processing.^16,17,25 Phenotypic markers indicating an increased phagocytic and antigen cross-presentation ability similar to DCs such as CD1C, CD1B, and CD207 have been identified in MDM population.¹¹ Notwithstanding, the phenotype of MDMs remains highly plastic with the increased expression of several pro-inflammatory markers including TSPO, ISG15, IFITM2, ANXA2, IRF7, S100A11, LGALS, and IL1b,¹⁷ and the cytokines CCL8, CCL13, CCL17, and CCL18.¹¹

Alternative clustering has identified an APOE-expressing TAM subset that shows the strong enrichment of antigen processing and presentation, which also expresses immunomodulatory molecules such as the complement C1Q chains, SPP1, and HLA-related molecules. An S100A8 expressing subset characterized by the high expression of S100A family genes CXCL8 and FCN1, and the low expression of HLA-related genes has also been reported, which resemble myeloid-derived suppressor cells with the enrichment of cytokine-mediated signaling and response to IL1.²⁶

Functional contribution of TAMs along the metastasis cascade

.—TAMs have been implicated as integral players in the pathogenesis of BrM (Figure 2). Already early during metastatic dissemination, endothelial-derived DKK1 release triggered by lung cancer-derived exosomes was shown to induce a protumorigenic phenotype in microglia, contributing to the establishment of the premetastatic niche.²⁷ In agreement, breast cancer cell-derived ANXA1 steered microglial migration prior to metastatic establishment and ANXA1 induced microglial activation via STAT3.²⁸ Aiding early metastatic dissemination to the brain, microglia activation by IL6 and CSF1 facilitates tumor cell BBB transmigration.^25,29 In experimental melanoma models, intravital imaging established an involvement of activated TAMs in metastasis formation by disrupting BBB integrity via MMP3.³⁰ Depletion of TAMs with an immunosuppressive phenotype by mannosylated clodronate liposomes in animal models further abrogated breast cancer BrM burden during early disease stages.³¹

TAMs play an established role in brain invasion, particularly by virtue of their interactions with tumor cells during the invasion/metastasis cascade. Upon extravasation of singular tumor cells, microglia get recruited, resulting in direct contact and communication with the cancer cells.^21,25 Microglia have been shown to aid brain invasion in a WNT pathway-dependent manner, which was reversible by either WNT inhibition or LPS-elicited activation.³² Furthermore, tumor cell invasion-inflicted perturbations in the brain structure, which are associated with microglial damage response, can spark enhanced tumor invasion in a CXCR4-dependent manner.¹⁵ This microenvironmental determination of tumor cell invasion is in line with an overall lack of association between primary tumor type and invasion patterns.³³

Finally, several mechanisms facilitating colonization also involve TAMs. Microglia expressing CD74 were shown to promote metastatic growth via midkine (MDK) secretion.³⁴ PI3K was additionally identified as a central regulator of TAM-mediated colonization, PI3K inhibition altering metastatic growth.³⁵ Moreover, cathepsin S released by MDMs was implicated in enhancing brain colonization.³⁶ Colonization was further assisted by tumor-derived IL6, which abrogated microglial inflammation via the JAK2/STAT3 axis in NSCLC BrM. Elevated serum levels of IL6 correlate with BrM occurrence and survival.²⁹ A trilateral interaction has been characterized, in which astrocytic exosomes inhibit BrM PTEN expression, thereby inducing alterations in chemokine secretion by tumor cells, which in turn enhance the attraction of bone marrow-derived macrophages to the metastatic site.³⁷ In addition, increased myeloid cell content has been correlated with immunosuppression, increased metastatic cell proliferation, as well as therapeutic failure.³⁸ Similarly, the intratumoral density of CD163+ TAMs correlates with shorter survival.³⁹ In line, microglia-conditioned media exhibit tumor-supportive properties in vitro.⁴⁰

As part of the innate immune system with intrinsic antitumor capacity, phagocytosis of early extravasated tumor cells has been observed in TAMs.¹⁹ Notably, the phagocytic capacity of TAMs was lost during disease progression, accompanied by TAM activation and MDM influx. Comparable phagocytic and tumorolytic capacity have been described using intravital imaging.⁴¹ LPS-induced activation of both microglia and MDMs could enable tumor cell killing in vivo and in vitro, underlining the overall potential for an effective antitumor immunity of TAMs.⁴⁰ Interestingly, tamoxifen treatment in an animal model of breast cancer ameliorated BrM by shifting microglia toward a more phagocytic phenotype.⁴²

Beyond the well-described involvement of TAMs in BrM establishment, their functional importance in orchestrating the niche immunity via multifaceted interactions with other immune cells, encompassing not only pro-tumor but also antitumor functions, has been a controversial field of debate. Pointing toward antitumor immunity orchestrated by microglial TAMs in concert with other immune cells, animals lacking microglia presented with enhanced tumor progression, paralleled by decreased natural killer (NK) cell and T-cell responses.²³ In line, microglial markers correlate with increased survival in breast cancer BrM.²³ Furthermore, TAM-secreted chemokines such as CCL3/4 are central for lymphocyte recruitment.²⁴ Interestingly, the elevated abundance of microglia and monocytes correlated with reduced local recurrence and leptomeningeal involvement in melanoma BrM.¹² This finding indicates not only a role for microglia in antitumor immunity but also a potential protective effect of monocyte infiltration, supporting the notion that naive monocytes still retain immunocompetence, unlike the tumor-shaped MDMs.⁷

Conversely, evidence points toward TAM involvement in shaping the immunosuppressive nature of the brain-MME via secretion of factors such as galectin-1,²⁴ Cystatin C,⁴³ or the production of immunosuppressive metabolites via IDO upregulation.⁴⁴ These changes are believed to limit antitumor immunity by altering T-cell–mediated immunity: increasing CD8+ T-cell exclusion, reducing T-cell activation, and increasing immune checkpoint expression²⁴ while facilitating regulatory T-cell (T-reg) attraction.⁴⁵ Microglial CX3CR1 expression loss, as typically observed in BrM, can induce a microglial support to BrM outgrowth by recruiting CD68+ myeloid cells with high expression of checkpoint markers PD-L1, VISTA, which hamper T-cell–mediated immunity.¹⁷ This aligns with transcriptomic studies implicating TAMs, particularly MDMs, in immunosuppression and altered T-cell function.^16,46 Conversely, metastatic-derived factors such as neurotrophin-3⁴⁷ and amyloid-β⁴⁸ have been demonstrated to foster immunosuppression by limiting microglial activation and phagocytic capacity.

Altogether, the evidence presented suggests a multifaceted involvement of TAMs in pro- and antitumor immunity. While limited evidence points towards an effective antitumor immune response by resident microglia during the initial steps of BrM, the educational influence of tumor-intrinsic and environmental factors potentially hijacks microglial-mediated immune signaling to facilitate metastasis outgrowth. While available evidence indicates a more tumorigenic function of infiltrating MDMs, insufficient data exist to clearly delineate the immunological contribution.

Neutrophils

Neutrophils represent 50%–70% of the myeloid-derived circulating leukocytes in human blood and are involved in innate immune response against invading pathogens. They also represent most of the inflammatory cells in solid tumors with a high intratumor density and are increasingly identified as key drivers of tumor progression by dictating TME composition and tumor vascularization.⁴⁹ Neutrophil increase can be observed in an elevated neutrophil-to-lymphocyte ratio in peripheral blood,⁵⁰ which has been associated with poorer survival postsurgical resection.⁵¹ Phenotypically, BrM-associated neutrophils show an induction of immune and inflammatory pathways including TNFα signaling, whereas ROS production is reduced.⁵⁰ BrM-associated neutrophils are characterized by high expression of infiltration and transmigration markers such as ITGA3, CD177, CD15, and CD11b.^11,50 In line with this immune-suppressive phenotype, neutrophils exhibit high expression of PD-L1, which was most strongly detected in vicinity to PD1+ CD8+ T cells, implicating a role for neutrophils in hampering antitumor immunity.⁵⁰ Immunosuppressive ARG1- and PD-L1-expressing neutrophils were shown to drive metastatic progression following attraction by tumor-derived G-CSF.⁵² Beyond immune-related functions, neutrophils have been implicated in angiogenesis, particularly localizing close to blood vessels within the tumor mass.^50,53 During early steps of BrM, the primary tumor induces astrocyte-mediated inflammation, which in turn causes an attraction of neutrophils to the metastatic niche.⁵⁴ Furthermore, granulocytes have been implicated in metastatic seeding following their attraction to the premetastatic niche by elevated levels of S100A8/9.⁵⁵ BrM-associated neutrophils also show elevated expression of S100A9,⁵⁰ potentially instigating a self-reinforcing cycle of neutrophil attraction. Intratumoral neutrophils express markers of recruitment of anti-inflammatory neutrophils (CXCL17, ADORA2A, and MET).^11,50 Based on these observations, strategies aiming at targeting tumor neutrophil infiltration by pharmacologically blocking chemokine signaling (eg, blocking IL8 and CXCL1)⁴⁹ or neutrophil-derived substances are currently viewed as valuable therapeutic options.

Dendritic cells

Acting as professional antigen-presenting cells, DCs are pivotal players initiating and regulating T-cell–elicited antitumor immune responses. A cluster of CD163 and CD14 expressing DCs was identified, which exhibits elevated chemokine expression, suggesting a role for T-cell recruitment.⁵⁶ Similarly, a cluster of type 3 DCs involved in T-cell activation and MHC expression was revealed in melanoma BrM.⁵⁷ Further exploration of patient samples of BrM of diverse tumor origin revealed a prevalence of type 2 DCs that subclustered into 2 fractions with mutually exclusive expression levels of either migratory receptors or antigen presentation.²⁶

Lymphoid Compartment

The brain contains a limited number of lymphocytic cells under physiological conditions, except for circumscribed functions of patrolling lymphocytes in the CNS border such as meninges and choroid plexus.⁵⁸ In stark contrast, lymphocytes represent the most prevalent immune cell type in the brain-MME, constituting 35%–65% of all cells.^10,11 The predictive value of tumor-associated lymphocyte abundance for patient prognosis is still disputed; however, the elevated presence of some lymphocyte subsets, particularly CD8+ T cells and cytotoxic lymphocytes, is associated with patient prognosis.⁵⁹ Conversely, progression under treatment was associated with lower lymphocyte density.⁶⁰

T Cells

T-cells constitute the majority of BrM-associated lymphocytes,^10,11 containing the largest phenotypic diversity of all immune cells in the brain-MME.²⁶ Blood vessels in the vicinity of metastatic cells express endothelial adhesion molecules such as VCAM1 and ICAM1 in different BrM entities,⁶¹ which aligns with reports of increasing abundance of both CD8+ and CD4+ T-cell subsets with increasing tumor grade.¹⁰

Infiltrating T-cells exhibit a mixed composition, consisting of both CD8 and CD4-expressing populations. Studies report a predominance of CD4+ T cells, with a high abundance of T-regs,^10,62 spanning a phenotypic spectrum from effector to central memory T cells.²⁶ Some reports describe the high accumulation of naïve and memory T cells, including both CD8+ effector memory and central memory T cells.^10,20,26 Notably, effector memory T-cells constitute the majority of CD8+ T cells, revealing a cytotoxic phenotype.^20,26 When tracking T-cell receptor (TCR) diversity, TCR clonal expansion occurred in cytotoxic T-cell fractions showing potential tumor reactivity, and singular TCR clones predominantly found in naïve T cells and T-regs.²⁰ The T-cell compartment in BrM is characterized by an activated and exhausted phenotype.¹⁰ Beyond extensive PD1 expression, other checkpoint molecules such as TIM3, CTLA4, TIGIT, and LAG3 are abundantly expressed.⁷ Interestingly, when investigating the organizational dimension underlying T-cell diversity, the activation state was revealed to orchestrate the phenotypic heterogeneity.²⁶ Similar to phenotypic changes observed in TAMs, changes in the activation state of T-cell subsets were paralleled by metabolic profiles.²⁶

Throughout BrM, dynamic interactions of inflammatory TAMs and T-cells have been reported. Preclinical evidence points towards an adaptation of an immunosuppressive phenotype of T cells upon contact with microglia.⁷ In contrast, T cells have recently been suggested to be responsible for inducing antigen presentation in tumor-associated microglia that are else restricted to interferon response.²³ Interestingly, TAMs isolated from patients identified to possess T-cell clones with potential antitumor capacity were shown to highly express both these two programs alongside strong T-cell recruitment capacity.⁶³

CD8 + T-cells

.—In line with the overall immunosuppressive nature of the T-cell compartment in BrM, CD8+ T-cells exhibit an exhausted signature, expressing PD1 and other co-inhibitory molecules such as CTLA4 and TIM3,^25,60 with only little overlap with peripheral CD8+ T-cells. Beyond the PD1+ cluster, a memory-like CD8+ subcluster was revealed.⁶⁰ These CD8+ subclusters also show spatially defined distribution based on antigen specificity, with exhausted CD8+ T-cells being most prevalent in the tumor core, whereas CD8+ memory T-cells were mainly present in the stroma-rich tumor margins.⁶⁰ Other studies have clustered infiltrating CD8+ T-cells phenotypically into a conventional cytotoxic and an interferon (IFN) signaling expressing cluster.⁷ Notably, a significant proportion of tumor-infiltrating CD8+ T-cells was characterized as reactive to nontumor, microbial antigens rather than tumor-specific ones.⁶⁰ In support, Wischnewski et al. reported that T-cells with low tumor-reactivity showed larger overlap with peripheral populations than tumor-antigen–specific CD8+ T-cells. Furthermore, a population of TOX-expressing CD8+ T-cells has recently been identified in melanoma BrM, which exhibits lower immune checkpoint marker expression compared with matched primary tumor samples.⁶⁴ Finally, several CD8+ subsets not only express markers of inhibition and activation but also co-express stimulatory receptors and proliferative markers.¹⁰

CD4 + T-cells

.—In general, a lower ratio of CD4+ T-cells and T-regs compared with CD8+ T-cells correlates with better prognosis.³⁹ CD4+ T-cells in BrM have been reported to have a hyporesponsive and anergic phenotype.¹¹ Several CD4+ T-cell subclusters have been identified in BrM, namely type 1 T helper-, Th17-, cytotoxic-, IFN-, and T-regs.²⁶ Showing expansion in both tumors and peripheral blood, T-regs have been revealed to induce immunosuppression and coincide with BrM progression.⁶⁵ Finally, γδ T-cells clustered into 2 main subtypes: IL17-producing and cytotoxic cells,⁷ and higher γδ T-cells abundance negative correlates with patient survival.³⁹

B-Cells

Evidence regarding the role of BrM-associated B-cells is scarce. In general, infiltration of B-cells is significantly lower than T-cells across cancers of origin.⁶⁰ High levels of intratumoral plasma cells expressing CD138 have been associated with increased overall survival in patients following resection and radiation.⁶⁶ In preclinical models, PD1 inhibition increased B-cell proliferation in BrM.⁶⁷ This is substantiated by evidence indicating a higher abundance of B- and plasma cells in melanoma patients that underwent immunotherapy.⁵⁷ Moreover, autoantibody levels have been shown to predict BrM.⁶⁸ Altogether, these changes may suggest a relevant role for B-cell activity during BrM progression and therapy response; however, the functional significance of these changes remains to be elucidated in BrM.

NK Cells and Other Emerging Lymphoid Subsets

NK cells harbor the unique ability to eliminate tumor cells without prior sensitization, positioning them a highly potent component of anticancer immunity. So far, little is known about the role of NK cells in the brain-MME, although animal models have indicated a functional relevance of NK cells because an increase in BrM burden upon NK cell depletion could be observed.⁶⁹ Latent disseminated cancer cells were shown to hamper NK cell-mediated immunity via downregulation of activators in a WNT-signaling–dependent matter.⁶⁹

There is growing interest in further lymphoid subsets such as mucosa-associated invariant T (MAIT) cells and innate lymphoid cells (ILCs). Although antigen presentation of microglia to MAIT-cells was documented,⁷⁰ current evidence on the presence of MAIT-cells in brain neoplasia is largely confined to primary brain tumors.⁷¹ There is only limited data on the role of ILCs in BrM, although a study in mice lacking ILC2 showed an increase in CTCs detected in the brain.⁷² Altogether, while the relevance of these lymphoid subsets in the TME is increasingly recognized, their specific roles in brain-MME remain to be elucidated.

Factors Influencing Immune Variation Illuminated by Differences between Brain-MME and Primary TME

Immune Distinctions between Brain-MME and Primary TME

The brain-MME differs substantially from primary brain-TME such as glioblastoma,^11,12 revealing unique BrM-specific features beyond the sole localization in the brain. While primary brain-TME feature high proportions of microglial TAMs, BrM-TAMs largely consist of MDMs.^10,53 Remarkably, microglia in BrM can acquire MDM features.¹¹ While notable divergence in microglial phenotypes are observed between primary and metastatic brain tumors, more phenotypic overlap exists in MDMs.¹¹ MDMs adopt a microglia-like phenotype independently of the primary tumor entity,⁹ likely reflecting adaptation to the localization in the brain. However, monocyte-to-macrophage transition is not homogeneous and is dictated by specific tumor entity.¹⁰ Increased vicinity of TAMs to T-cells is also observed in BrM compared with primary brain tumors,¹¹ theoretically allowing for both enhanced antitumor immunity and immunosuppressive activity by TAMs. Neutrophil infiltration and density is elevated in BrM,^50,56 as is plasma cell abundance.¹⁰ Moreover, T-cell abundance in the brain-MME exceeds that of primary brain-TME, which is limited to sparse perivascular accumulation.⁶⁹ TCR repertoires in BrM show higher interregional consistency, similar to peripheral TCR repertoires.⁷² Simultaneously, T-regs accounting for 10% of all CD4+ T-cells constitute a higher proportion of the brain-MME lymphoid compartment compared with brain-TME.^10,60 Finally, limited evidence in patient samples suggests that NK cells in the vicinity of BrM form a CD56 expressing, uniquely activated and faster proliferating cluster compared with NK cells in primary brain tumors.¹⁰

Besides brain-MME and brain-TME differences, the immune landscape of BrM also contrasts with that of peripheral primary tumors that metastasize to the CNS (peripheral-TME). The immune contrast between brain-MME and peripheral-TME is evident in the lymphoid compartment. CD8+ T-cell numbers are generally diminished in BrM compared with their corresponding primary tumors.⁴⁴ Notably, a reduced diversity of T-cell clones has been reported in BrM compared with primary lung cancer despite the higher mutational burden found in metastases.⁷³ An increased proportion of NK cells with diminished CD56 expression has been described in BrM compared with primary NSCLC, suggesting an impaired NK population in the metastatic niche.⁷⁴ Striking differences in B-cell abundance among different cancer entities have been described. Unlike the comparable B-cell compartment in primary and metastatic lung adenocarcinoma, metastases originating from breast cancer showed a stark reduction compared with primary tumors.⁷⁵ Great variation in the myeloid compartment such as elevated neutrophil infiltration and density in BrM has also been reported.^59,60 Comparing TAM subsets in brain-MME and peripheral-TME to elucidate further organ-specific immune interactions will offer insights for new therapeutic strategies. While important, a detailed analysis of MDMs and the resident macrophages (eg, microglia, alveolar, mammary, and dermal) is beyond the scope of this brain-focused review. Understanding the immune landscapes across TMEs as well as various MMEs holds promise for clarifying immune dynamics in BrM and guiding treatments.

Determinants and Modulators of the Immune Composition and Phenotype in BrM

The pronounced differences between brain-MME and primary TME highlight that the brain niche mainly acts as a modulator of the MME, while metastatic tumor cells primarily dictate the immune interactions. Single-cell analyses across patients and primary tumor types in BrM have revealed consistent stromal composition but significant variation in the immune compartment,²⁶ further indicating that metastatic cells from different primary tumor entity are key determinants of the brain-MME composition and phenotype. Immune cell frequency varies across a spectrum encompassing low infiltration levels in breast cancer, higher in lung cancer and strongest in melanoma BrM.¹⁰ Concomitantly, BrM stemming from breast cancer were shown to have a higher presence of TAMs, while higher infiltration of lymphocytes was shown in lung cancer (T-cells) and melanoma (B-cells).²⁶ Correlating with a better prognosis,³⁹ increased CD8+ T-cell abundance was most commonly found in melanoma BrM.¹² Beyond abundance, infiltration patterns also differ between primary tumor types. Stronger T-cell infiltration at tumor margins than cores has been reported in melanoma,⁶² while others report predominant stromal infiltration in NSCLC BrM.⁶⁸ Immune variance in the brain-MME can be determined by specific mutational profiles of metastatic cells. In lung cancer BrM, TP53 mutation has been shown to be associated with a more proliferative transcriptional signature of immune cells and a more immunosuppressive myeloid compartment, particularly TAMs.⁷⁶ Hypermutated breast cancer BrM, on the other hand, were characterized by a highly inflammatory brain-MME enriched with TAMs, neutrophils, and CD8+ T-cells.⁷⁶

In addition to the described effects of brain localization and primary tumor entity, further interpatient variance has been described, reporting preferential accumulation of TAMs in the peritumoral region,⁷⁷ in the tumor mass,²¹ or both.⁴⁰ CD8+ T-cells infiltration can range from heavy infiltration to almost complete absence.²⁰ The composition of the immune brain-MME has also been related to features such as patient age, with increased neutrophil, DC, and CD4+ T-cell abundance having been reported in younger patients.¹² Finally, previous treatment is believed to contribute to the observed interpatient variance.^57,60 Therapeutic interventions such as radiation in late disease stage can transiently shrink the myeloid compartment and shift the TAM composition back to a higher fraction of resident microglia due to increased vulnerability of MDMs.^18,57

Targeting the Immune Microenvironment in BrM

Given the lack of adequate therapeutic options for BrM and the dire prognosis of most patients, a pressing need for new therapeutic options exists. Considering the central role of the immune microenvironment in BrM, brain-MME-targeted therapies appear to be a promising approach, and therapeutic interventions targeting TAMs and T-cells have been tested in preclinical and clinical settings.

Targeting TAMs

The dual role of TAMs as both mediators of protumorigenic effects and potential conveyors of antitumor immunity renders them promising therapeutic targets. This is particularly true considering TAM phenotypic plasticity and their intrametastatic localization in vicinity to T-cells.¹¹ In general, two strategies of targeting TAMs have been tested, TAM depletion and TAM reprogramming. TAM depletion by CSF1R inhibitor has been proposed to effectively prevent BrM establishment in a melanoma mouse model.³⁰ However, TAM depletion by CSF1R did not achieve sustained tumor control due to the compensatory upregulation of CSF2R-mediated TAM activation in a mouse model of breast cancer. Targeting the downstream signaling target of CSF2R, STAT5, concomitantly with CSF1R inhibition leads to continued tumor control.²⁵ CSF1R blockade induced a TAM phenotype with similar expression patterns to neurodegenerative disorders, highlighting the detrimental effects of TAM targeting therapies.²⁵ In animal models of several primary tumors, systemic administration of the TLR9 agonist CPG-C triggered microglia to phagocyte, successfully preventing the establishment of BrM.⁴¹ Targeting anti-inflammatory TAMs by conditional knock-out of KLF4, which governs the STAT6 pathway, hampered BrM growth in a breast cancer mouse model, similar to pharmacological STAT6 inhibition.⁷⁸ Activation of STING in TAMs, unlike promising evidence indicating anticancer capacity in models of glioblastoma,⁷⁹ has not been shown to effectively counter BrM.⁸⁰ Conversely, STING activation has even been linked to increased occurrence of BrM.⁸¹

Finally, in vivo inhibition of the macrophage migration inhibitory factor (MIF), a crucial chemokine for tissue recruitment of TAMs, could synergize with radiotherapy against NSCLC BrM. Blocking MIF signaling via CD74 inhibition could reverse the effect of NSCLC-derived MIF, prompting microglia towards a more migratory, phagocytic phenotype in vitro.⁸²

T-Cell–Based Immunotherapy

Over recent years, classifications characterizing the immunological phenotype of tumors to predict response to immunotherapy have emerged, categorizing tumors on a spectrum from immunologically “hot” to “cold.”⁸³ Primary brain tumors with limited immune infiltration, such as glioblastoma, are defined as “cold” tumors. In contrast, BrM could be quantified as rather “hot” tumors due to the extensive presence of infiltrating lymphocytes and preexisting antitumor immunity, indicated by PD-L1 expression. Consequently, while “cold” tumors such as glioblastoma may benefit from the local introduction of tumor-targeting CAR-T-cells,⁸⁴ there is a strong rationale for harnessing T-cell immunity with immunotherapies in BrM.

Owing to the high expression of PD1 and PD-L1 in melanoma and breast cancer BrM,²⁶ clinical trials have recently focused on exploring the efficacy of immune checkpoint blockade in BrM, including CTLA4 and PD1 inhibition and their combinations.

A recent phase 2 study reported the investigation of the PD1 antibody pembrolizumab in BrM from a diverse set of untreated and pretreated primary tumors including breast cancer, lung cancer, melanoma, and sarcoma. Intracranial response was observed in around 40% of patients, with selected patients demonstrating survival exceeding 2 years. However, adverse effects were observed in more than 50% of patients.⁸⁵ Besides, intracranial response to the combination of PD1 inhibitor nivolumab and CTLA4 inhibitor ipilimumab in melanoma BrM was shown to range from 46% to 57% depending on the clinical study but was associated with adverse events in more than half of all patients.^86,87 A recent trial of nivolumab and ipilimumab in NSCLC revealed increased 5-year as well as intracranial progression-free survival compared with chemotherapy.⁸⁸ Encouragingly, a phases II–III trial combining the LAG3-targeting antibody relatlimab and nivolumab demonstrated increased time to melanoma BrM development compared with nivolumab only in a phases II–III trial.⁸⁹ Preclinical evidence further suggests a synergistic effect of targeted therapies with checkpoint inhibition, combination of the CDK4/6 inhibitor abemaciclib sensitizing melanoma BrM animal models to PD1 inhibition.⁶⁷

Besides tumor-intrinsic and microenvironmental features, differences in immunotherapy response between BrM and primary brain tumors might also be influenced by the presence of a primary tumor. In a preclinical model of melanoma BrM, the presence of a peripheral tumor was required for intracranial response by increasing CD8+ T-cell trafficking to the BrM.⁹⁰ There is a possibility that neoantigens expressed by the primary tumor cells may have primed T-cells in the periphery prior to BrM development, potentially enabling these primed T-cells to target the BrM expressing the same neo-antigen.⁹⁰ Immunotherapy seems to benefit from neo-antigen abundance at the time of treatment, as suggested by improved outcomes of neo-adjuvant over adjuvant immune checkpoint blockade in glioblastoma.⁹¹ However, low-reactive T-cells have showed greater overlap with peripheral populations than tumor-specific CD8+ T-cells.⁶⁰ Therefore, the presence of a primary tumor may influence T-cells beyond direct antigenic priming, supporting the rationale for neo-adjuvant timing of immunotherapy.

The Potential of Combinatory Regimens

Given the significant interindividual variance in the brain-MME immunophenotype, some strongly T-cell infiltrated BrM clearly present as “hot” tumors indicative of a potential response to immunotherapy, other BrM can display the sign of immune exclusion and overall immunosuppressed brain-MME. This variability underscores the rationale for combinatory therapies modulating the brain-MME in addition to strengthening T-cell response.

Existing well-established therapies such as chemoradiation already target cells in the brain-MME beyond tumor cell-intrinsic effects. For instance, radiation has been shown to reduce immunosuppression and increase cytotoxic T-cell numbers.⁴⁶ Interestingly, the transient change in immune microenvironmental phenotype was subsequently reported to be countered by the increase of immunosuppressive PD-L1 + myeloid cells.⁴⁶

Chemotherapy in combination with PD-L1 antibody atezolizumab showed intracranial response rates of 45% in NSCLC BrM,⁹² while combinatory treatment consisting of pembrolizumab and chemotherapy of NSCLC patients with BrM revealed a survival benefit of 12 months compared with chemotherapy alone.⁹³

Combining stereotactic radiosurgery with ipilimumab has further been reported to increase survival in melanoma BrM.⁹⁴ In line, a recent meta-analysis of both whole brain radiation as well as stereotactic radiosurgery in combination with immunotherapy in melanoma BrM revealed increased survival compared with radiation alone.⁹⁵ This reported synergy between radiation and immunotherapy likely relates to anticancer effects in nonirradiated brain regions, termed the abscopal effect.⁹⁶ Underlying mechanisms include radiation-triggered cytokine release and elevated tumor antigen presentation that activates microglia and other immune cells, as well as the elimination of immunosuppressive cells in the microenvironment by radiation.⁹⁷ Finally, the altered BBB permeability postradiation and subsequent influx of peripheral immune cells as well as efflux of brain-restricted antigens into systemic circulation have been proposed to underlie the abscopal effect.⁹⁸

While chemotherapy combinations oftentimes aim to achieve highest concurrent doses for maximum anticancer efficacy, the fine-tuned nature of the immune responses elicited by radiation necessitates careful considerations regarding scheduling of radiation and immunotherapy combinatory treatment.⁹⁷ A trend towards improved prognosis for immunotherapy administered in the first 4 weeks after radiation has been demonstrated.⁹⁵ Similarly, improved prognosis for patients with melanoma BrM treated with ipilimumab following compared with preceding stereotactic radiosurgery has been shown.⁹⁹

Combinatory treatment of targeted therapy and immunotherapy in BRAF-mutated melanoma BrM has also been reported to benefit from employing immunotherapy as a follow-up treatment.¹⁰⁰ Currently, an on-going clinical trial is investigating combining NK cell therapy and chemotherapy for brain metastatic melanoma (NCT05588453) (Table 1). While early results of TAM and T-cell–directed therapies are promising, investigation of combined approaches that aim at rewiring the brain-MME is still pending. In this regard, novel approaches combining tumor-targeted, pro-tumoral TME-targeted, and immunotherapies could be considered for BrM.

Conclusion

A wealth of evidence has indicated the significant role of the immune microenvironmental niche in the pathogenesis of BrM, arising from the complex interactions between tumor cells and surrounding immune cells. While the majority of previous studies considered tumor cells as the central determinants of brain-MME cellular composition and immune condition, the advent of high-throughput single-cell analyses has provided unprecedented insights into the signaling programs within both malignant cells and the bystander immune microenvironment in the metastatic brain tumors. It is increasingly evident that specific tissue conditions can also decisively predefine the susceptibility of a tissue for metastatic invasion as well as contribute significantly to the progression of a tumor. Thus, a thorough understanding of the signaling circuitries within the metastatic immune niche holds great promise to foster the development of novel therapeutic approaches as well as innovative combinatory treatment strategies for patient benefit.

Acknowledgement

All figures were generated with Biorender.com.

Conflict of interest statement. R.B. has received honoraries for lectures and advisory boards from AbbVie, Amgen, AstraZeneca, Bayer, BMS, Boehringer-Ingelheim, Illumina, Janssen, Lilly, Merck-Serono, MSD, Novartis, Qiagen, Pfizer, Roche, Targos MP Inc. R.B. is a co-founder and co-owner of Gnothis Inc./Stockholm (Sweden) and Timer Therapeutics/Freiburg (Germany). All other authors declare no competing interests.

Author contributions

Conceptualization: L.D.S., R.G., P.-H.N. Writing—original draft: L.D.S., P.-H.N. Writing—review and editing: L.D.S., A.F.v.S., S.T.J., M.T., K.-W.N., R.B., H.K., V.N., R.G., P.-H.N. Visualization: A.F.v.S., L.D.S., P.-H.N. Supervision: R.G., P.-H.N.

Funding

P.-H.N., R.B., and H.K. are funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) (grant SFB1530-455784452) and by the research network CANcer TARgeting (CANTAR) of the Ministry of Culture and Science of the State of North Rhine-Westphalia (MKW NRW).

References